Possibly the result of utilising a CDI module (CDI has been around in one form or another for the past 40 years). Do you know whether it used a CDI module?
That usage case pretty well describes an emergency petrol/gas powered generator which has rather narrowly defined operating parameters. In this case, you're either looking at a nominal 3000/3600 rpm on a small four stroke powered generator or else a restricted rpm range of around two to four thousand rpm (give or take) for a small 1 to 2 KW rated inverter generator intended to be kept sheltered from monsoon storm like conditions with a very short HT lead between coil and the single spark plug.
Damp ingress is unlikely to be a problem and the engine life between major services only a fraction of that of its spark plug even under operational conditions of extreme asymmetric spark discharge currents. In this case, a CDI module would be an overkill solution to a largely non- existent problem and a simple HV switching transistor would be the most pragmatic option.
My remark about the shittiness of the simple HV switching transistor analogue of the classic Kettering ignition system was made in the context of (off)road going vehicles using spark ignited IC engines where the operating conditions are far more arduous.
Getting back to your original question, presumably, you won't be using
12v ignition coils with a 24v backed constant current supply in the final production run and you're just asking about why your test setup should produce the phenomena you've been witnessing when the coil is subjected to more than 16 volts. For my part, I have no firm idea as to what's actually happening. I can try making a guess or two but they're not likely to be any better than anyone else's 'best guess' so I won't muddy the waters with untested theorising. I'll leave that to others more brave than I. :-)